Caching in dataflow-based environments

We propose that there are several advantages to integrating caching in the execution scheme of a dataflow-based environment. If each node has a cache to store its last result, the cache look-up overhead is eliminated, because each cache contains just one value. A node in a dataflow-based program must store its result anyway until its downstream neighbors are ready to execute. Additionally, there is no need to store previous arguments with the cached result for comparison. The validity of a node's cached value can be determined from the validity of the nodes on which it depends.